Research On Error Detection And Concealment Techniques In Video Communication

With the developing of electronic technology, computer science, telecommunication and theory of image compression, digital video communication can be found today in many application scenarios. Usually, the video data is compressed before transmission, and the compressed streams are sensitive to the channel error or network congestion. Unfortunately, an inherent problem with any communication system is that information may be altered or lost during transmission due to channel noise. This conflict presents a number of challenges for high-quality image and video transmission. Many techniques have been proposed to deal with transmission errors, which can be divided into three groups: error resilience encoding, interactive error compensation and error concealment (EC) by post processing. In this paper, we dive into the research of the error detection and concealment with effective novel algorithms propsed. The main research topics and contributions are listed below.Based on the mode-based error detection (MED) method, an error detection and localization technique using syntactic, semantic and statistical information in MPEG-4 is proposed. In the proposed approach, transmission errors within an MPEG-4 bitstream are detected by means of two successive procedures. Whether the video packet is corrupted or not is firstly determined by checking a set of error-checking conditions. Then the exact location of the first transmission error within the corrupted video packet is located by using a block-based checking procedure. Experimental results show the error detection ratio (EDR) of the proposed technique for different types of transmission error is usually more than 90%. Besides, the technique can increase the video quality efficiently by combining with EC method. Another advantage of the technique is the bit rate of video bitstream will not be increased, which is important for multimedia communication.An error concealment method for true 2-D wavelet codec based on set partitioning in Hierarchical trees (SPIHT) is proposed. For the lowest frequency wavelet coefficients, the EC scheme based data hiding is performed. The lowest frequency coefficients, which are taken as the hidden data, are extracted from the compressed bitstream, and embedded back into the same bitstream. The restored hidden data is used to conceal errors at the decoder. The damaged high frequency coefficients are predicted through linear interpolation based on intra-subbband or inter-subband correlation. Experimental results show that the proposed method achieves good and stable error resilience performance with minimal additional redundancy.Concerning inter4v mode employed widely in MPEG-4 video, an adaptive temporal error concealment approach based on macroblock (MB) mode for MPEG-4 video sequences is proposed. It can selectively interpolate one or four motion vectors (MVs) for the missing MB according to the estimated MB coding mode. Performance of the proposed approach is compared with the existing schemes with multiple testing sequences at different bit error rates (BERs). Experimental results show that the proposed algorithm can mask the impairments caused by transmission errors much more efficiently than 0 MV and average MV methods by consuming a little more time for different BER. It yields an acceptable image quality close to the selective motion vector matching (SMVM) algorithm, while taking fewer than half of cycles of operations. The proposed concealment scheme is suitable for low complexity video real-time implementations.A hybrid error concealment algorithm for MPEG-4 video is proposed. For the missing MBs in intra-coded frame and inter-coded frame, two different classifiers based on the neighbor are firstly presented. Then the categorised MB is reconstructed by a suitable EC method selected adaptively by the category-dependent criteria reflecting the properties of the block content. Simulation results show that the subjective visual quality and quantative qualities of image concealed by the hybrid algorithm yield greater improvement than any single EC method, while the computational complexity of the proposed algorithm is satifactory.Some recent error concealment algorithms for shape information in MPEG-4 are also reviewed. Besides, the performance of several representative spatial and temporal error concealment methods for shape information are tested and compared by multiple MPEG-4 testing sequcences.The proposed algorithms are applicable to other MPEG, H.26x and JPEG images with no or a bit of modifications.